Do scholars' collaborative tendencies impact the quality of their publications? A generalized propensity score matching analysis

Recently, the research on the cooperation relationship between authors has received widespread attention. However, existing studies still have the following limitations: 1) They mainly study the impact of author collaboration patterns by correlation analysis without considering the existence of confounding factors. 2) Methods based on causal analysis primarily focus on exploring the impact of different cooperation models, while less considering the author's tendency to participate. 3) Previous studies fail to incorporate the structural attributes of the authors' cooperation network into covariates, which may lead to confounding bias. To overcome the above limitations, we further explore the causal effect of authors' participation levels on the quality of their publications by leveraging the Generalized Propensity Score Matching (GPSM) method. Moreover, to alleviate the influence of the structural features in the authors' cooperation network, we then take the typical structural features as covariates, preventing us from reaching incorrect conclusions caused by the variable bias. We conduct extensive experiments on a real-world dataset (collected from the Web of Science (WoS) core collection), and from the experimental results, we find that authors having different involvement tendencies usually have publications with different qualities. Specifically, we observe an “inverted U-shaped” curve on authors' participation tendencies. That is, the quality of papers first rises and then decreases with the increase of authors' participation tendencies, which means that researchers who excessively collaborate with others actually experience a decrease in average paper quality.